Ricursive's Ripple-65B: $32M-trained open model claims US open-weight TerminalBench SOTA
annadgoldie · x · 2026-11-05
Anna Goldie amplified RicursiveAI's release of Ripple-65B Preview, a dense 65B LLM pretrained from scratch.
- $32M total R&D; claims US open-weight SOTA on agentic coding benchmark TerminalBench 2.1, beating much larger models like Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B.
- Runs locally on a MacBook Pro or DGX Spark; architecture designed around the hardware it trains and runs on, feeding a "better models design better chips, better chips train better models" flywheel.
- An internal variant already powers Ricursive's AI chip design platform. Nathan Lambert praised it as proof small focused teams can build solid models, intellectually extending the OLMo lineage.
Related event: Ricursive releases Ripple-65B, claims open-weight agentic coding SOTA(14 posts)→
More from coding & agent
- Sentry CEO shows AI agent shipping a Restream plugin and asking about API scope risks — zeeg · 2026-11-05
- Open-source TrnRT engine serves GLM-5.3-Flash on one Trainium2 box at 3.57M tokens/min — ycombinator · 2026-11-05
- Raindrop engineer makes the case for building your own internal Slack agent — zeeg · 2026-11-05
- Genkit Go, Eino, ADK Go, tRPC-Agent-Go, LangChainGo compared with real code — rseroter · 2026-11-05
- How do you connect cloud AI agents to tools behind a customer's firewall? — JoleneAtCynapsa · 2026-11-05
- Elixir guide for OO programmers by Bruce Park goes free online with an AI tutor — burkov · 2026-11-05